Re: [HACKERS] proposal: session server side variables

On 6 January 2017 at 08:44, Jim Nasby <Jim.Nasby@bluetreble.com> wrote:

>>        (1) private/public visibility (as Oracle does with package vars).
>>            this point is enough to implement the presented use case.

Agreed.

>>        (2) typing (casting is a pain)

We already have typed GUCs and allow them to be user-defined. See
DefineCustomBoolVariable, DefineCustomIntVariable, etc.

What we lack is a way to declare and use typed dynamically
user-defined GUCs at runtime without a C extension.

We also lack user interface to load and store values without going via
their text representation; there's no current_setting_int4 etc.

So if we allow for now the idea that we'd extend the GUC model to do
this (which I'm not at all sure is a good thing) ... it's possible.

>>        (5) have some "permanent" GUC type declarations (maybe editing the
>>            config file does that already, by the way?)

We have that, but it currently requires a C extension.

-- Craig Ringer                   http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services